Acute Gastrointestinal Hemorrhage


3 The presentation of GI bleeding can range from mild asymptomatic bleeding to overt GI bleeding. UGI bleeding typically presents with hematemesis (the vomiting of blood) or melena, whereas LGI bleeding typically presents with hematochezia. Melena is a black, tarry stool resulting from the degradation of blood by enteric bacteria. It may occur with the loss of as little as 50 to 200 mL of blood.20,21 Bleeding from the small intestine or right colon may also appear black if it has remained in the GI tract for more than 12 to 14 hours.22 Hematochezia is the passage of bright red blood, maroon-colored blood, or blood clots from the rectum. However, massive UGI hemorrhage can cause hematochezia in as many as 11% of patients, but this is typically associated with hemodynamic instability.23 Patients with acute GI bleeding may present with the hemodynamic consequences of hemorrhage including light-headedness, dizziness, orthostatic syncope or near syncope, shortness of breath, or palpitations from tachycardia.

Table 65-2 Differential Diagnosis of Acute Lower Gastrointestinal Hemorrhage by Anatomic Site

Figure 65-1. A: The relative frequency of the most common causes of upper gastrointestinal hemorrhage in the United States. These data represent the percentage of patients with each of these causes of UGI hemorrhage for 482 patients in a survey of the members of the American College of Gastroenterology published by Peura et al.1 in 1997. These data are very similar to that reported by Vreeburg in a multi-institutional study of 951 patients sustaining a UGI hemorrhage in the hospitals in and surrounding Amsterdam.4 B: The relative frequency of the most common causes of lower gastrointestinal hemorrhage. These data, reported by Lingenfelser and Ell5 are the percentage of patients with each of these causes of lower GI hemorrhage in 912 patients collected in five studies from Europe, the Orient, and the United States1,710

The medical history and physical examination provide important clues of the etiology of the patient’s hemorrhage and the potential risk to the patient’s life. The occurrence of melena after several days of worsening epigastric or upper abdominal pain suggests peptic ulcer disease; whereas hematemesis or melena following vomiting or retching strongly suggests a Mallory–Weiss tear. Massive, painless UGI hemorrhage in a patient with cirrhosis suggests bleeding from gastroesophageal varices, although other etiologies including peptic ulcer disease or a Mallory–Weiss tear must also be considered. The medical history should elicit the presence of risk factors for GI hemorrhage alluded to in the previous paragraphs and in Table 65-3 .

A systematic physical examination will document the magnitude of bleeding and the patient’s ability to compensate. Massive hemorrhage is associated with signs and symptoms of hypovolemic shock, including cool, clammy, mottled skin, tachycardia, tachypnea, flat jugular veins, oliguria, and perhaps hypotension. These responses may be altered by advanced age, concomitant medical problems, and particular medications. Physical examination should also document evidence of cirrhosis and portal hypertension (i.e., ascites, spider angiomas, hepatosplenomegaly, palmar erythema, and large hemorrhoidal veins). A rectal examination may demonstrate bright red blood or melena. The clinical scenario alone will usually not localize the location of the bleeding, so other diagnostic studies are often required to identify the cause and site of bleeding.


4 Most patients (approximately 80%) suffering from GI hemorrhage will stop bleeding spontaneously. Those who do not stop or those who rebleed are at particularly high risk to suffer an in-hospital complication, require operative control of their hemorrhage, or die. Several classification systems have been developed to separate patients with low risk of complications from those with a high risk of complications due to acute UGI and LGI hemorrhage. These systems have also been used to stratify those patients who may be safely managed as an outpatient from those requiring in-hospital care.24 The BLEED classification system addresses both UGI and LGI hemorrhage and consists of the following parameters: ongoing bleeding, low systolic blood pressure, elevated prothrombin time, erratic mental status, and unstable comorbid disease. Patients with at least one BLEED criterion are more likely to suffer in-hospital complications from UGI bleeding (31% vs. 4%) or LGI bleeding (38% vs. 12%) than are patients with no criteria.25

Table 65-3 Characteristics of Individuals at an Increased Risk of Developing Acute Gastrointestinal Bleeding

Table 65-4 Rockall Risk Scoring System and Rates of Rebleeding and Mortality

Prognostic systems for UGI hemorrhage have been more widely adopted than those for LGI hemorrhage. One widely used system is the Rockall score for assessing the risk of death and rebleeding in patients with UGI hemorrhage (Table 65-4). Using this model, Rockall et al. found that rebleeding occurred in less than 5% of patients and mortality was virtually zero (0% to 0.2%) in patients with scores of 0 to 2. In contrast, a fourth to nearly one-half of patients with a Rockall score of 5 to 8+ rebled; the mortality rate for these patients was 11% to 41%. In this study, rebleeding significantly affected the likelihood of death, particularly for patients with intermediate scores of 3 or 4 and 5 to 7 in which there was a three- to fivefold increase in mortality rates.26

The Rockall classification has been widely accepted as accurate and importantly has been externally validated.2730 However, the full classification scheme requires endoscopic assessment. An alternative scoring system – the Glasgow–Blatchford bleeding score (GBS) – is based on clear and readily available clinical and laboratory indices without the need for endoscopy Table 65-5. It was designed to predict need for clinical intervention due to UGI hemorrhage. This scoring system has been subjected to multi-institutional trials and found to be at least as effective as and possibly more accurate than the Rockall system.31,32 These authors suggested that patients with a score of 0 can be safely managed as outpatients.

Conversely, patients with LGI bleeding typically present with less hemodynamic instability. Factors that is indicative of a severe lower GIB thus necessitating more urgent intervention include:

Heart rate >100/min

Initial systolic blood pressure 115 mm Hg or less

History of syncope

Nontender abdominal examination

Bleeding per rectum during the first 4 hours of evaluation

History of aspirin use

Charlson Comorbidity Index score of more than 2.33


Upon presentation, two large-bore intravenous lines should be placed in peripheral veins and intravascular volume resuscitation begun with an isotonic saline solution. Most patients stop bleeding spontaneously, and crystalloid volume resuscitation is all that is required. Blood is drawn for type and crossmatch, complete blood count with platelet count, electrolyte measurement, liver function tests, and coagulation profiles. It is important to emphasize that on presentation, the hematocrit or hemoglobin level may not accurately reflect the magnitude of acute blood loss. Estimates of the severity of hemorrhage must be based on clinical parameters.

The massively bleeding patient should receive packed red blood cells (RBCs) to restore intravascular volume and oxygen-carrying capacity. The decision to transfuse blood or blood products depends on the individual needs of the patient and the cause of the bleeding. The risks of the blood products (i.e., infection and allergic reactions) must be weighed against the risks of withholding transfusion (i.e., anemia, decreased oxygen-carrying capacity, coagulopathy). In general, blood products are used early in the management of patients with limited cardiac and pulmonary reserve (who are unable to withstand or compensate for an acute reduction in their systemic oxygen delivery) and those with lesions that are at particular high risk for continued or recurrent hemorrhage (e.g., gastroesophageal varices).

Careful hemodynamic monitoring of these potentially critically ill patients is vital to successful management. There are no clear recommendations, but it seems reasonable for those patients who are actively bleeding and those who have recently sustained significant hemorrhage to be admitted to an intensive care unit for close monitoring of hemodynamic parameters and evidence of continued or recurring hemorrhage. The presence of significant underlying illnesses, such as cardiac, renal, hepatic, or pulmonary insufficiency, may necessitate noninvasive monitoring or invasive cardiac monitoring with central venous and arterial catheters. The information gained from these devices allows cardiac performance to be optimized during intravascular volume replacement. The placement of a urinary catheter and frequent monitoring of heart rate, blood pressure, urine output, and mental status are the minimum necessary to monitor patients who have suffered GI hemorrhage. The importance of prompt, adequate resuscitation and diligent observation cannot be overemphasized as the cornerstone for managing these potentially mortally ill patients.

Table 65-5 Glasgow–Blatchford Bleeding Score (GBS)


5 After the restoration of circulating blood volume, the next step is to identify the source of bleeding so that definitive therapy may be instituted. If the patient presents with hematemesis, localization of the bleeding to the esophagus, stomach, or duodenum is relatively straightforward and esophagogastroduodenoscopy (EGD) should be performed promptly to identify the source of bleeding. When blood or coffee-ground guaiac-positive material is present in the gastric aspirate, EGD will likely define the site of bleeding. Bright red blood per rectum strongly suggests a lower GI source of bleeding unless the patient is hemodynamically unstable in which case the hemorrhage may originate from a source proximal to the ligament of Treitz. A general algorithm for evaluating patients with acute UGI and LGI hemorrhage is presented in Algorithm 65-1.

Algorithm 65-1. Diagnostic steps in the evaluation of gastrointestinal hemorrhage.

Table 65-6 Predictors of Persistent or Recurrent Bleeding in Patients with Nonvariceal UGI Hemorrhage

Gastric Aspiration

Recent studies have called into question the routine placement of nasogastric tubes (NGTs) in all patients with suspected GI bleeding. In patients with hematemesis, gastric aspiration is not necessary to obtain a diagnosis of a UGI bleed. However, it can provide useful information regarding the rate of hemorrhage. Knowing the degree to which a patient is bleeding may help guide clinical decision making, such as the need to initiate octreotide, transfuse blood emergently, urgently perform an EGD, or determine whether or not ICU monitoring is necessary.34

In the absence of hematemesis, aspiration of gastric fluid after the placement of a NGT may be used to distinguish between a UGI and LGI source of bleeding. Two studies with more than 700 patients found that the presence of blood in the gastric aspirate was a good indicator of a UGI source; however, its absence was unreliable in predicting the presence or absence of a UGI source.35,36 In one study of 220 patients with UGI sources of bleeding, the sensitivity, specificity, and accuracy of the nasogastric aspirate was 42% (95% confidence interval [CI] = 32, 51%), 91% (95% CI = 83, 95%), and 66% (95% CI = 59, 72%), respectively.36 While a clear aspirate does not rule out a UGI source of bleeding, aspiration of blood confirms it. Additionally, placing an NGT may help remove blood from the gastric lumen to improve visualization once the EGD is performed. However, NGT placement can be associated with pain, aspiration, and pneumothorax.37 Additionally, other less invasive methods may indicate a UGI source (black stool, BUN/creatinine ratio >30, and age younger than 50 years). Due to these factors, NGT placement prior to EGD should not be routine but instead individualized to each patient.



EGD will identify the site of bleeding in about 95% of cases of UGI bleeding and is the initial diagnostic study for patients suspected of bleeding from the esophagus, stomach, or duodenum.38 The sensitivity of the procedure is significantly enhanced when performed within the first 24 hours of presentation.4 A systematic review of the literature found that early endoscopy (i.e., performed within 24 hours of admission) was associated with a decreased transfusion requirement and decreased length of stay.39 A prospective, randomized trial found that early endoscopy allowed the triaging of 46% of patients to outpatient care without any adverse effects.40 Previous consensus guidelines and several cohort studies have related various endoscopic stigmata of recent or active hemorrhage to a heightened risk of rebleeding or continued bleeding.4144 Laine and Peterson41 analyzed data from 37 prospective trials in which patients with bleeding ulcers did not receive endoscopic therapy; they found that the rate of further bleeding was less than 5% for those patients with a clean ulcer base and increased to 10% for patients with a flat spot, 22% for those with an adherent clot, 43% for those with a nonbleeding visible vessel, and 55% for those with active bleeding. Endoscopic features predictive of persistent or recurrent bleeding and mortality are shown in Tables 65-6, 65-7, and 65-8. In addition to these ulcer-specific factors, endoscopy allows identification of lesions with a high risk of continued hemorrhage and mortality (i.e., gastroesophageal varices), and those with a low risk (e.g., Mallory–Weiss tears). The efficacy of endoscopy-based modalities to control UGI hemorrhage is discussed in subsequent sections.


Although the efficacy of colonoscopy in determining the cause of occult GI bleeding is undisputed, historically its role in the evaluation of patients with acute LGI bleeding was less well agreed upon. Recently however, newer studies have established it as the procedure of choice for patients with suspected LGI bleeding despite its limitations.4548 Colonic lavage with a polyethylene glycol solution can be used to clear the lumen of clot and stool providing adequate visualization of the mucosa,49 but others have reported good visualization of the mucosa even in the absence of mechanical bowel preparation.50

Studies have shown a diagnosis is made in 74% to 100% of patients with an LGI bleed undergoing colonoscopy, and a pool analysis of six recent studies found a composite yield of 91% for colonoscopy.47,48,51,52 A meta-analysis examined the role of colonoscopy as the primary diagnostic modality for patients with acute lower GI bleeding and found that 69% (range 48% to 90%) of urgent colonoscopies identified a source or a presumptive source of bleeding.53 Even in the setting of unprepped bowel, urgent colonoscopy has been shown to identify bleeding colon and distal ileal lesions in 82 of 85 patients (97%).50 Stigmata of recent hemorrhage for LGI bleeding are similar to those of UGI lesions and include an actively bleeding site, a nonbleeding visible vessel, and an adherent clot; these findings have been associated with continued hemorrhage and therefore the need for urgent colectomy.7,54 Jensen et al.45 reported that 25% to 50% of patients with any of these three factors continued to bleed or rebled and ultimately required urgent colectomy. Others have found colonoscopy to be less accurate in the diagnosis of LGI bleeding, for example, Al Qahtani et al.55 reported a series of 136 patients in which colonoscopy identified only 45% of the sources of bleeding.


For those patients who present with hematochezia in whom the initial EGD and colonoscopy is nondiagnostic, repeating these studies before evaluating the small intestine is warranted given the very small frequency in which the bleeding originates from the small intestine (1%). Repeating the EGD and colonoscopy when the patient is better resuscitated will often detect lesions such as ulcers or vascular ectasias that were obscured by blood at the initial endoscopy or the vasoconstriction of the GI mucosa that accompanies hemorrhagic shock.

Endoscopy of the small bowel with an enteroscope or a pediatric colonoscope will allow inspection of the proximal 60 cm of the jejunum56 and the use of a long videoenteroscope may allow visualization of 100 to 150 cm of intestine beyond the ligament of Treitz.57 Jensen et al.58 in an experience with more than 200 patients with obscure sources of GI bleeding reported success in identifying the etiology in 79% of instances using enteroscopy. In their experience, vascular ectasias and postbulbar ulcers were the most common causes of obscure GI bleeding.

Table 65-7 Predictors of Mortality in Patients with Nonvariceal UGI Hemorrhage

Intraoperative enteroscopy using a combination of push enteroscopes per os and per rectum or via enterotomy can allow examination of the entire small bowel. While the endoscopist manipulates the scope, the surgeon manually advances the bowel over the endoscope. After the bowel is telescoped onto the endoscope it is slowly withdrawn while the endoscopist examines the mucosal lumen and the surgeon watches the transilluminated bowel wall. While this technique can be effective, it is limited by its invasive nature.59

Double Balloon Enteroscopy

This technique utilizes a long enteroscope and a long overtube. Both the overtube and the enteroscope have balloons at the end. When the balloon of the enteroscope is inflated it “grabs” the mucosal surface and allows advancement of the overtube whose balloon is deflated. The overtube balloon is then inflated while the enteroscope balloon is deflated. The enteroscope is then advanced while the inflated overtube balloon grips the mucosa. Using these alternate inflation–deflation cycles, long distance advancement of the enteroscope has been achieved.2 In one U.S. multicenter study, the average distance achieved was 360 cm with a diagnosis made in 43% of cases.60 In some cases lesions may be seen that are missed by other techniques.61

Wireless Capsule Endoscopy

Imaging of the small intestine is also possible with wireless capsule endoscopy (CE) which consists of a battery, light source, imaging-capturing system, and transmitter. This capsule endoscope is 11 × 30 mm and is moved solely by peristalsis. This system captures and sends usually two images per second (newer versions can send more frames per second) for about 8 hours to an ultra–high-frequency band radiotelemetry unit worn by the patient. In 85% of cases, the capsule reaches the cecum within this time frame.62 The location of the capsule is suggested by the strength of the signal. Several studies have shown high diagnostic yields using this technique and found it to be at least equivalent to and in some studies superior to double balloon enteroscopy in patients with obscure GI bleeding.63 It has also been shown to have a higher diagnostic yield with comparable outcomes when compared to angiography.64 The main risk of CE is capsule retention. One difficulty with CE is the very large amount of data to be reviewed. For an 8-hour study, 50,000 to 60,000 images are generated, which takes approximately 40 to 60 minutes to review.

Selective Visceral Arteriography

Selective visceral arteriography is primarily useful in patients with UGI or LGI bleeding in whom endoscopy cannot be performed or has been unsuccessful in determining the site of ongoing, rapid hemorrhage. Successful angiographic identification of the source of bleeding occurs in 27% to 86% of instances and depends primarily upon the presence of active arterial bleeding at the time of the study (Fig. 65-2).55,65 The extravasation of contrast may be detected if the patient is bleeding at rates greater than 0.5 to 1 mL/ min66; this correlates clinically with the requirement for continuous volume infusion to maintain hemodynamic stability. Pennoyer et al.,67 however, were unable to identify any clinical parameters (including tachycardia, numbers of transfusions, or orthostatic hypotension) that could increase the diagnostic yield of selective angiography, including scintigraphy demonstrating ongoing bleeding. However, a study by Hammond et al.68 found an early (within 2 minutes) positive scintigraphy has a 60% positive predictive value for a positive angiogram. Additionally, it has been shown that a positive scintigram increases the likelihood of a positive angiogram from 22% to 53%.69 Several groups have used heparin, vasodilators, or thrombolytics to improve the diagnostic yield of arteriography in patients with nondiagnostic studies.70,71 Mernagh et al.72 found that the administration of heparin intravenously for 24 hours increased the diagnostic yield of visceral angiography from 33% (6 of 18) to 67% (12 of 18). Others have found that the intra-arterial infusion of a vasodilator, heparin, and/or urokinase (a thrombolytic) failed to identify the source of bleeding in 5 of 7 patients.73 It should be noted that these provocative techniques are not commonly employed.

Figure 65-2. Selective celiac arteriography with injection into the common hepatic artery in a patient bleeding from a duodenal diverticulum. Extravasation of contrast from a branch of the gastroduodenal artery can be seen (arrow).

One major advantage of visceral arteriography is its therapeutic potential. Transcatheter embolization of bleeding vessels was first reported in the early 1970s.74,75 Modern instruments allow superselective catheterization of terminal vessels allowing satisfactory embolization with less risk for ischemic complications. Further discussion of therapeutic use can be found below.

Abdominal Scintigraphy

Abdominal scintigraphy with 99mtechnetium (Tc)-labeled RBCs lacks the spatial resolution and diagnostic precision of angiography and endoscopy; however, it is of most value in detecting intermittently bleeding lesions, those with very low rates of hemorrhage such as vascular malformations, and in evaluation of LGI bleeding due to the lower sensitivity of endoscopy within the LGI tract as compared to the UGI tract. Abdominal scintigraphy utilizing 99mTc-RBCs has been shown to be the most sensitive examination due to its ability to detect bleeding rates as low as 0.04 to 0.1 mL/min.76,77 In a review of seven retrospective studies with nearly 400 patients, the median diagnostic accuracy of scintigraphy was 82% (range, 52% to 95%). However, in this review, 99mTc-RBCs was incorrect in 5% to 48% of instances (median 18%).65 Suzman et al.78 summarized 20 retrospective studies containing 804 positive studies and reported a false-positive rate of 19%. The large variation in these studies results from differences in scan timing, technical skills, and expertise. Additionally, precise localization of the site of bleeding may be complicated by the rapid distribution of isotope throughout the intestine by peristalsis or by accumulation in the right colon.79 More recent techniques of cine scintigraphy may improve the diagnostic accuracy (Fig. 65-3).80 One area where radionuclide scanning has a clear role is in the diagnosis of Meckel diverticulum. 99Tc-pertechnate is secreted by ectopic gastric mucosa in Meckel diverticula.81 This study should be considered early in the evaluation of young individuals with LGI bleeding.

Figure 65-3. Cine-99Tc erythrocyte scintigraphy showing extravasation of isotope in the right colon. Only a small portion of the image set is shown. Arrows point to accumulation of isotope in right colon. Bleeding was due to delayed hemorrhage following endoscopic polypectomy.

CT Scanning

CT scanning has been used to detect GI bleeding using a variety of specialized techniques.8284 Multidetector-row helical computed tomography (MDCT) is establishing itself as a rapid, noninvasive, and accurate diagnostic method in acute GIB. In arterial phase MDCT, active GI bleeding is defined as active contrast extravasation with a focal area of high attenuation within the bowel lumen.85 Yoon et al.86 presented the first prospective study evaluating the accuracy of MDCT and found an overall sensitivity of 90.9% for the detection of acute GI bleeding and specificity of 99%. Additional studies have supported its use.87,88 Limitation of MDCTs are radiation dose, contrast allergies, contrast nephropathy, and it being a purely diagnostic rather than therapeutic modality. However, it is fast, accurate, widely available, reproducible, noninvasive, and a positive study can subsequently guide interventionalists to directly perform time-saving super-selective angiograms of the bleeding site.85 In the future this may continue to become a more available and useful technique.


Upper GI Hemorrhage

Peptic Ulcer Disease

Peptic ulcer disease is the most common cause of acute UGI hemorrhage, accounting for nearly 40% of cases in most series although this proportion may be decreasing.89 About 15% to 20% of patients with peptic ulcer disease experience bleeding during the course of their disease and as many as 20% of these patients will have bleeding as the initial manifestation. Hemorrhage is the principal cause of death from peptic ulcer disease and has replaced intractable pain as the most frequent indication for surgery. Complications of peptic ulcer disease occur more commonly in older patients who often have coexisting medical problems, which profoundly influence their risk of morbidity and mortality.

Duodenal ulcers occur slightly more frequently than gastric ulcers. Penetration of the ulcer through the posterior wall of the duodenal bulb is associated with erosion into the gastroduodenal artery or one of its branches, resulting in brisk hemorrhage. Patients may present with hematemesis of bright red blood and clots or with melena alone. Between 80% and 90% of patients stop bleeding spontaneously during the initial stages of therapy with volume resuscitation.

In general, patients with gastric ulcers tend to be older and have coexisting medical problems that increase morbidity and mortality compared with duodenal ulcers. Bleeding may occur from any site in the stomach, although ulcers occurring at the incisura are most common. At this site, involvement of the branches of the left gastric artery may result in brisk, if not torrential, hemorrhage. The clinical presentation of patients bleeding from gastric ulcers is similar to that of duodenal ulcers, with hematemesis, melena, and hematochezia.

6 An important risk factor for the development of GI hemorrhage and gastroduodenal ulcer formation is the use of NSAIDs. NSAID use has been associated with a continuum of mucosal injury ranging from small acute mucosal hemorrhages to large chronic ulcers. It has been estimated that 10% to 15% of regular NSAID users have chronic gastric ulcers.90 Symptoms correlate poorly with the degree of mucosal injury since as many as 20% of ulcers penetrating the muscularis are asymptomatic.91 Case control and cohort studies have suggested that NSAIDs are associated with a relative risk of GI hemorrhage and ulceration ranging from about 2 to 9.1.92 Ketorolac, in particular, has been associated with a high risk of GI bleeding (relative risk approaches 25).93 The risk of NSAID-associated complications is highest in patients with a history of UGI bleeding, the elderly, those patients taking oral anticoagulants,94 or corticosteroids. Patients with a prior history of peptic ulcer disease also appear to be at increased risk of NSAID-associated GI hemorrhage and tend to have a significantly worse outcome when compared to individuals not using NSAIDs.95

The tremendous frequency with which NSAIDs are used by the elderly underscores the magnitude of this problem. Individuals with a history of NSAID-induced hemorrhage may benefit from the prostaglandin E1 analog, misoprostol, which has been shown to prevent NSAID-induced gastric erosions and ulcers.96 Histamine (H2)-receptor blockers (ranitidine and cimetidine) are effective in preventing NSAID-induced duodenal ulcers, but appear to have little effect on the occurrence of gastric lesions.91,97 Proton pump inhibitors (PPIs) have also been shown to be protective in patients on NSAIDs, with greater efficacy than H2 blockers.98 NSAIDs are also associated with lower GI bleeding, including lesions not generally considered related to NSAID-induced ulcers such as diverticulosis.99,100 Selective COX-2 inhibitors have been marketed as being safer than nonselective NSAIDs, although some of these have been withdrawn from the U.S. market for other safety concerns. It appears that these selective inhibitors cause fewer UGI problems overall than traditional NSAIDs, the main benefit being fewer uncomplicated ulcers. Various studies have shown no decrease in complicated events, including clinically significant bleeding episodes.101,102 However, in a Cochrane systematic review of the GI safety of COX-2 inhibitors, COX-2 inhibitors produced significantly fewer gastroduodenal ulcers (relative risk, 0.26; 95% CI = 0.23–0.30) and ulcer complications (relative risk, 0.39; 95% CI = 0.31–0.50), as well as fewer withdrawals caused by GI symptoms when compared to nonselective NSAIDs.103

Medical Treatment

7 Once bleeding from the UGI tract is confirmed, treatment with a PPI should be initiated. Acute use of a PPI has been shown in several studies to decrease rebleeding.43,104107 Although the Scottish Intercollegiate Guidelines Network (SIGN) recommended withholding PPI therapy until after endoscopy,108,109 in our opinion the balance of evidence would suggest no harm and perhaps a benefit from early PPI use. An 80-mg intravenous bolus of omeprazole or pantoprazole followed by an infusion at 8 mg/hr produces the most reliable acid suppression.110 All patients should be tested for Helicobacter pylori (H. pylori) infection and treated if found. Treatment of the infection significantly reduces the recurrence of hemorrhage when compared to no treatment or chronic antisecretory treatment alone.111 Interestingly H. pylori infection may be less common in patients with bleeding ulcers than in those with nonbleeding ulcers.112

Endoscopic Treatment

The endoscopic appearance of a bleeding ulcer has important prognostic and therefore therapeutic implications, as alluded to in Tables 65-6 and 65-7. A modification of the system employed by Forrest et al.113 is shown in Table 65-8. In this system, category I findings are indicative of active bleeding while category II findings provide evidence of recent hemorrhage. In general, only actively bleeding ulcers (i.e., Forrest category I lesions) are treated endoscopically.

Table 65-8 Forrest Classification of Endoscopic Appearance of Bleeding Ulcers

A variety of endoscopic techniques are available to arrest hemorrhage from bleeding ulcers. The precise method of treatment is less important than the correct selection of patients and the experience of the endoscopist. Examples of methods typically include mechanical, thermal, or injection.

Mechanical ligation of bleeding vessels can be achieved with endoscopic ligation (banding), endoscopic clipping, or endoloop ligation.114116 Of these three methods, endoscopic clipping is the method most commonly employed for bleeding ulcers.

Heater probes and monopolar and bipolar electrocoagulation probes can also effectively control UGI hemorrhage. Monopolar probes apply high-frequency electrical current to the tissue, resulting in localized heating to 100°C and sealing of the bleeding vessel by coagulation necrosis of the surrounding tissue and vessel wall. Multipolar electrocoagulation (bicap) probes consist of three equally spaced pairs of bipolar microelectrodes. This orientation of electrodes allows coagulation of tissue from tangential approaches and eliminates some of the disadvantages of the monopolar probe, such as the unpredictable depth of thermal injury, adherence of tissue, and clot dislodgement. Direct thermal coagulation of a bleeding point can also be produced by applying a heater probe, consisting of an aluminum tip coated with Teflon. The tip is rapidly heated to 250°C by an inner coil. The tip can be irrigated with a water jet to prevent accumulation of debris and clot. Heat conducted from the probe produces tissue coagulation to a depth of 1 to 5 mm.

Injection of epinephrine to induce vasoconstriction has been used successfully to control acutely bleeding ulcers, particularly as an adjunct to electrocautery or mechanical hemostasis with clips. A meta-analysis of 15 studies concluded that injection alone was inferior to either clips alone, or clips plus injection.117 This study showed no difference between clips and thermocoagulation.

Although not commonly employed, the injection of sclerosants has been well described as a method of treating esophageal varices and has been used for controlling nonvariceal bleeding. Sodium morrhuate and ethanolamine oleate are most commonly used to treat esophageal varices, whereas ethanol and polidocanol are most commonly used for nonvariceal sites. These agents act by thrombosing bleeding vessels and causing necrosis and subsequent fibrosis of surrounding tissue. Clinical experience with sclerosants has been similar to that obtained with electrocoagulation. In one large multicenter study of 332 actively bleeding patients or patients with stigmata of recent hemorrhage who underwent injection of 98% alcohol around the lesions, less than 1% continued to bleed, 6% rebled, and only 3% required emergency operative intervention.118

A meta-analysis of 25 randomized trials of endoscopic therapy for bleeding ulcers concluded that endoscopic treatment methods have a beneficial effect on survival by reducing the rate of recurrent hemorrhage. This analysis suggested that endoscopic therapy results in a relative reduction of 69% in recurrent bleeding, 62% in emergent surgery, and 30% in mortality rate, with the greatest benefit seen in actively bleeding ulcers and ulcers with nonbleeding visible vessels.119 The effectiveness of early aggressive endoscopic diagnosis and treatment is further supported by a report of 562 patients bleeding from a variety of causes of whom only 2.5% required emergency operations to control hemorrhage.120 Several other more recent studies have confirmed the critical role of endoscopy and control of UGI hemorrhage.43,121,122 For most patients with evidence of persistent bleeding, a second attempt at endoscopic hemostasis should be attempted because this can provide up to 75% of patients with durable hemostasis. Exceptions may include patients with ulcers greater than 2 cm in diameter and those who have hypotension associated with a rebleeding episode, since such patients may be at an increased risk for failure of repeat endoscopic hemostasis.104,121,123

Operative Treatment

The successful use of endoscopic therapies has relegated operative procedures to a rescue role for those cases in which endoscopy is unsuccessful in arresting hemorrhage. Numerous studies have attempted to identify those patients at greatest risk of continued or recurrent bleeding. Of the many factors examined, those associated with the highest risk of rebleeding included patients in hypovolemic shock during the initial endoscopy, ulcers greater than 2 cm in diameter, and endoscopic stigmata of recent or ongoing hemorrhage (Forrest type I and II lesions).123 Many studies have demonstrated the ability of endoscopy to identify those patients at greatest risk of rebleeding. In one review, the presence of active bleeding was associated with a 90% to 100% chance of continued or recurrent hemorrhage. A nonbleeding visible vessel had a 40% to 50% chance, adherent clot 20% to 30%, oozing without visible vessel 10%, flat spot 5% to 10%, and clean-based ulcer 1% to 2%.124 Even in those patients who rebleed following initial endoscopic therapy, two-thirds may be successfully retreated endoscopically thus avoiding operative intervention.90 Factors that must be considered in decisions regarding the timing of operative intervention include the magnitude of the initial (or recurrent) hemorrhage, the physiologic ability of the patient to withstand continued or recurrent hemorrhage, and the likelihood of recurrent or continued hemorrhage. It is generally accepted that elderly patients and those with significant concurrent medical problems should undergo operative intervention earlier during the course of the hemorrhage since these individuals will poorly tolerate continued bleeding, recurrent hypotension, and repeated transfusions.

The type of operation depends on the pathology encountered. For bleeding gastric ulcers, the operation of choice depends on the patient’s condition and location of the ulcer. For favorably located ulcers, excision of the ulcer with closure of the gastrotomy will suffice. If the ulcer is unfavorably located, for example, near the gastroesophageal junction, simple oversewing of the vessel through the base of the ulcer may adequately control bleeding.125 If a gastric ulcer is left in situ, follow-up endoscopy is necessary 4 to 8 weeks later to either confirm healing or obtain tissue to rule out malignancy. Extensive gastric resections such as antrectomy, subtotal or total gastrectomy are generally not performed in these unstable patients.

For patients bleeding from duodenal ulcers, the type of surgery performed has changed since the development of PPIs and treatment of H. pylori. Formerly, truncal vagotomy, pyloroplasty, and oversewing of the bleeding vessel were the most widely used operation. However, the trend is now more toward direct ligation of the bleeding vessel through the duodenotomy.126 When performed, ligation should incorporate the gastroduodenal artery proximal and distal to the ulcer as well as the transverse pancreatic artery. Despite the above noted trend, a recent study has shown vagotomy and drainage may be superior to simple oversewing of the vessel.127

Angiographic Embolization

Of late, much has been written about the use of transcatheter embolization for peptic ulcer disease. In general, angiographic embolization is used after failure of endoscopic treatment in patients who cannot or will not undergo surgery. One large review (including nonulcer UGI indications) showed a high technical success rate (i.e., localization of bleeding and deployment of the embolic agent) but a clinical success rate of only 51%.128 Other reviews have shown higher success rates ranging from 69% to 100% technical success rate and 63% to 97% clinical success rate.129 Significant ischemic complications can occur however. While this technique has utility in select patients, it should not be considered as a first-line treatment option for bleeding peptic ulcers, but may have a role as a safe alternative to surgery for ulcers refractory to endoscopic treatment.

Stress Gastritis

Although studies in the 1960s and 1970s demonstrated acute erosions of the gastric mucosa in as many as 60% to 100% of critically ill patients, the incidence has markedly decreased over the past four decades. Factors postulated to have been important in this phenomenon include (1) the widespread use of prophylactic gastric alkalinization; (2) improvements in the ability to detect and treat sepsis; (3) improvements in the ability to monitor and correct hemodynamic instability; and (4) the ability to provide adequate nutritional support of critically ill patients.

In general, stress gastritis is characterized by the appearance of multiple superficial gastric ulcerations within 12 to 14 hours of an acute injury. These lesions, initially localized to the fundus and body of the stomach, later involve the entire gastric surface. Patients at greatest risk include those with sepsis, major burns, severe trauma, and critically ill patients with a coagulopathy and respiratory insufficiency.130 In this setting, the disease appears to represent the gastric component of the multiorgan failure syndrome.

The pathogenesis of this disease is discussed in detail in Chapter 45. The primary defect is in the protective processes that maintain the integrity of the gastric mucosal barrier. Although some gastric acid secretion is required for the development of stress gastritis, it is clear that the hypersecretion of acid is not the cause of mucosal injury. Altered gastric mucosal blood flow and impaired clearance of hydrogen ions from the mucosa appear to be of particular importance; therefore, the best method of prevention of stress gastritis is to prevent gastric ischemia and acid injury. Stress gastritis should be differentiated from the deep, often solitary ulcerations occurring in patients with severe central nervous system lesions (Cushing ulcers).

Generally hemorrhage is the only symptom that patients with stress gastritis experience. Overt bleeding is often heralded by the appearance of flecks of blood in the gastric aspirate. The superficial nature of the lesions makes perforation unlikely.

Prophylactic therapy is directed toward preventing hemorrhage, primarily by neutralizing gastric acid, augmenting mucosal defenses, and removing or preventing physiologic stress. The gastric pH should be maintained between 3.5 and 4.5.131 Antacids, H2-receptor antagonists, PPIs, and sucralfate have all been used to prevent stress gastritis. Alkalinization of the gastric contents is associated with oral and fecal flora colonization of the stomach and has raised concerns about an increased risk of nosocomial pneumonia. This concern prompted the use of sucralfate as a preferred prophylactic agent for some time instead of antacids or cimetidine.132 However, a prospective, randomized trial of 1,200 critically ill patients receiving either ranitidine or sucralfate for stress ulcer prophylaxis found that those patients receiving ranitidine had a lower bleeding rate (1.7%) than the sucralfate group (3.8%) but there was no difference in mortality or incidence of ventilator-assisted pneumonia.133 Since that time, studies have shown PPIs to be a safe and effective method of stress prophylaxis, but the number of studies with high-quality data is still lacking.

The success of these prophylactic measures has led to a dearth of recent experience in managing patients bleeding from stress gastritis. Based on early reports, attention to blood replacement, intravascular volume restoration, and correction of coagulation defects are associated with the cessation of hemorrhage in nearly 80% of cases and as such are the principal means of initial treatment. A variety of nonoperative techniques have been employed with variable success in arresting hemorrhage from stress gastritis including endoscopic and embolization techniques and the selective catheterization of the left gastric artery with continuous infusion of vasopressin.134

Based on these same early experiences, very few patients bleeding from erosive gastritis require operative intervention to arrest hemorrhage. A variety of surgical treatment options have been reported including vagotomy and pyloroplasty with oversewing of bleeding sites, vagotomy and hemigastrectomy, total gastrectomy, and gastric devascularization. The dilemma facing the surgeon is that these critically ill patients poorly tolerate extensive procedures, yet lesser operations often fail to control hemorrhage. Regardless of the operation performed, mortality risk depends on the underlying illness, particularly in the presence of multiple organ failure. Mortality rates between 30% and 60% are commonly quoted, with as many as one-fourth of the deaths resulting from continued hemorrhage. Rebleeding rates ranging from 25% to 61% have been reported. The combination of vagotomy, hemigastrectomy, and oversewing of bleeding points has been touted as more successful in these patients; however, rebleeding rates of 11% to 44% and operative mortality rates ranging from 33% to 63% have been associated with this procedure.135 More extensive operations, such as near total gastrectomy or total gastrectomy are associated with significant mortality although they successfully stop hemorrhage.

Gastroesophageal Varices

Cirrhosis is a leading cause of death in the United States and variceal hemorrhage is a common mode of death for these patients. About 30% of people with cirrhosis develop gastroesophageal varices; of these individuals, about 30% bleed as a result of the varices, usually within 1 to 2 years of diagnosis. Gastroesophageal varices are a significant cause of UGI hemorrhage, accounting for about 20% of such cases. Patients with bleeding gastroesophageal varices tend to have much higher rebleeding rates, transfusion requirements, lengths of hospitalization, and risk of death than do patients bleeding from nonvariceal causes.120,136

Although the basic tenets of resuscitation for massive variceal hemorrhage are similar to those for any cause of massive bleeding, intravenous volume resuscitation should be particularly judicious. The hyperaldosteronemic state of cirrhosis promotes sodium and water retention with aggravation of ascites and peripheral edema. Accurate blood replacement is imperative since overtransfusion may worsen portal hypertension and exacerbate hemorrhage. Invasive cardiac monitoring with Swan–Ganz catheterization may be particularly useful for guiding volume replacement. Coagulation deficits should be aggressively corrected by administering fresh frozen plasma. Thrombocytopenia, secondary to hypersplenism or dilution, should be treated promptly with pooled platelet transfusions. Sedatives are best avoided or used sparingly because cirrhosis impairs the liver’s ability to metabolize many of these drugs. Adequate prophylaxis for delirium tremens should be administered to alcoholics.

As with other sources of UGI hemorrhage, early endoscopy is imperative for successful diagnosis and therapy.137 The identification of varices alone is not adequate to incriminate them as the source of the hemorrhage since up to half of patients with cirrhosis bleed from a source other than varices. Furthermore, endoscopy may identify factors associated with a heightened risk of variceal hemorrhage such as the size and number of varices and the presence of red, blue, or other colored spots on the varix. The presence of gastric and duodenal varices and portal hypertensive changes in the gastric mucosa (portal gastropathy) will influence therapeutic decisions and prognosis.

Although vasopressin has commonly been used in the management of variceal hemorrhage, its use has been limited due to the potent vasoconstrictive properties causing cardiac and peripheral ischemia, arrhythmias, hypertension, and bowel ischemia. More recent reports suggest the superiority of somatostatin or its synthetic analog, octreotide. It is thought that octreotide causes splanchnic arteriolar vasoconstriction and reduces variceal and azygous vein flow with limited direct effects on portal pressure.138 Meta-analyses have shown that the infusion of somatostatin is more effective and safer than vasopressin in the pharmacologic control of variceal hemorrhage.139,140 Other studies have shown that somatostatin or octreotide can improve the results of sclerotherapy or endoscopic variceal ligation.141,142 Although neither somatostatin nor vasopressin plus nitroglycerin definitively treat the bleeding esophageal varices, these modalities may provide initial control of hemorrhage, reducing transfusion requirements and providing time for resuscitation before definitive treatment.

Another temporizing method used for massively bleeding patients is balloon tamponade using a Sengstaken–Blakemore tube or a Minnesota tube. These devices consist of a gastric tube with esophageal and gastric balloons. In the case of a Minnesota tube, a proximal esophageal lumen allows for the aspiration of swallowed secretions. Inflation of the gastric (and if required esophageal) balloons tamponade the bleeding varices, controlling hemorrhage in more than 80% of cases. Hemorrhage recurs in 25% to 50% of patients upon deflation of the balloons, thus limiting this technique to a temporizing role.143 The greatest value of these tubes is for arresting massive hemorrhage that has been unresponsive to other measures, allowing time for resuscitation and angiographic definition of the portal system before definitive treatment.

When used inappropriately, these tubes can be associated with significant morbidity and mortality. Complications occur in 4% to 9% of patients with the most frequent being aspiration pneumonitis. Measures to prevent pulmonary complications include endotracheal intubation before tube insertion and the placement of an esophageal tube to remove swallowed salivary secretions. Other significant complications include esophageal rupture or necrosis and airway occlusion during the attempted removal of an incompletely deflated gastric balloon. Because of the availability of endoscopy, these tubes are rarely used today.

8 Endoscopic sclerotherapy, endoscopic clipping or endoscopic variceal ligation (banding) have become the most widely used modalities for the initial definitive control of bleeding esophageal varices. Several studies have confirmed that these techniques arrest acute variceal hemorrhage in 90% to 95% of patients. In general, a patient bleeding from esophageal varices should undergo urgent pharmacologic therapy and banding of the varices at the time of the first emergency endoscopy. Sclerotherapy can be used if ligation is technically difficult.137 A single endoscopic treatment controls variceal bleeding in more than 70% of patients, and a second treatment increases the rate of control from 90% to 95%. Continued or recurrent hemorrhage after endoscopic treatment often requires emergency portal decompression either with transjugular intrahepatic portosystemic shunting (TIPS) or rarely surgery.144146 Following an initial episode of variceal hemorrhage several options are available for the prevention of further hemorrhage. These options, as well as the operative management of bleeding esophageal varices are discussed in detail in Chapter 59.

Mallory–Weiss Tears

The Mallory–Weiss syndrome is acute UGI hemorrhage that occurs after retching or vomiting. Mallory and Weiss described the laceration of the gastric cardia and postulated that violent emesis against an unrelaxed cardia was the mechanism of injury. They were able to produce similar mucosal tears in cadavers by forcing gastric contents against an occluded gastroesophageal junction.147 The typical patient is an alcoholic who begins to retch and vomit after an alcohol binge although this syndrome may also be found in nonalcoholics with bouts of emesis. Initially the vomitus consists of gastric contents without blood and subsequently the patient develops hematemesis and/or melena. Overall, these lesions account for about 5% to 10% of patients with UGI bleeding.120,136,148

The initial management of these patients is similar to that of patients bleeding from other sources of UGI hemorrhage and includes volume resuscitation, gastric lavage, and decompression. Most patients with Mallory–Weiss tears stop bleeding spontaneously, either before treatment or after these early measures. Once bleeding has stopped, rebleeding is rare.

In patients who continue to bleed despite these maneuvers, nonoperative and operative therapeutic options are available. Nonoperative management, consisting of endoscopic electrocoagulation, banding or injection therapy, has been successfully applied to these lesions.149 In cases not amenable to endoscopic therapy, operative management consists of oversewing the laceration through an anterior longitudinal gastrotomy in the middle third of the stomach.


Although the passage of maroon or bright red blood per rectum may occur in the presence of a massive UGI hemorrhage, this finding most commonly indicates a source distal to the ligament of Treitz. The absence of blood in bilious nasogastric lavage further supports a distal location of hemorrhage. Although numerous potential causes of LGI hemorrhage are possible (Table 65-2), colonic diverticulosis and colitis are by far the most common. Small bowel sources and other colonic pathology such as colon cancer are relatively unusual causes of acute GI hemorrhage.

Diagnostic Approach

The most important question to answer when presented with a patient with LGI hemorrhage is not, “What is bleeding?” but rather, “Where is the bleeding?” The common causes of colonic hemorrhage are mucosal in nature, not palpable, and not visible from the serosal surface of the bowel. Therefore, it is imperative that the surgeon makes every effort to localize the source of bleeding preoperatively since it is usually impossible to locate intraoperatively. If the surgeon waits to attempt localization until it is clear that surgical treatment will be required it may be too late. Colonoscopy, tagged RBC scanning and/or angiography should be obtained as early as possible after presentation.

After determination that the bleeding is likely from an LGI source, it is important to first exclude anorectal causes of hemorrhage, such as hemorrhoidal bleeding. At this point, colonoscopy is usually performed as it can rule out anorectal causes of bleeding, but also help determine non-anorectal causes of hemorrhage. In the authors’ experience, despite studies showing high value, colonoscopy in an actively bleeding patient with an unprepped colon is seldom therapeutic. However, it can be of significant diagnostic benefit. It is usually fairly easy to determine when bleeding is from a neoplasm, and noting the location helps guide the surgeon in decision making. Conversely in AVMs and diverticular bleeding, locating the actual source of bleeding is often unsuccessful due to the presence of blood and stool. However, noting the extent of blood in the colon can be helpful to guide future surgical decision making. For example, if blood is only noted in the sigmoid colon and rectum, this suggests bleeding is from a distal source and a sigmoid colectomy may suffice if bleeding persists. Alternatively if blood is noted throughout the entire colon, the location of the bleed is unclear and a total colectomy may be necessary. The terminal ileum should be intubated as well. While a small amount of blood can reflux into the terminal ileum, if an extensive amount of blood is noted then a small bowel source should be considered and the surgeon should be hesitant about performing a total colectomy if bleeding persists.

Alternatively, a tagged RBC scan can be obtained as soon as practical, or in occasional cases of massive hemorrhage, conventional angiography or MDCT. If the tagged RBC scan localizes a bleeding site, the patient is then sent for angiography to attempt to embolize the bleeding vessel. If the tagged RBC scan does not demonstrate bleeding, then it is probable that bleeding has stopped. One may then proceed to colonoscopy after adequate mechanical bowel preparation.

If the bleeding is rapid enough to warrant emergency surgery, bleeding can usually be demonstrated by tagged RBC scan, MDCT, or angiography. It should be the very rare patient who will need to proceed to surgery for LGI hemorrhage with failed preoperative localization. In these cases a careful search for small bowel bleeding sources, possibly including intraoperative small bowel endoscopy, should be performed.

Colonic Diverticulosis

9 In Western society, the prevalence of colonic diverticula increases with age such that about 60% of people in their seventh decade of life are affected and the incidence increases roughly 1% per year. Only about 20% of patients with diverticulosis have symptoms attributable to these lesions and less than 5% experience hemorrhage.150 Hemorrhage from diverticular disease is most often massive, associated with hematochezia, and accompanied by varying degrees of hemorrhagic shock. Classically, patients present with a sudden occurrence of mild lower abdominal discomfort, rectal urgency, and the subsequent passage of a large bloody stool. Because the colon can contain large volumes of blood, neither the volume nor the frequency of bloody stools is a reliable guide to the rate of hemorrhage. Despite the massive nature of hemorrhage, most patients with diverticular disease stop bleeding spontaneously. Most series report bleeding ceases spontaneously in 80% of cases, although up to 25% to 50% can rebleed. Of those who rebled, less than one-fourth required surgery.151

Bleeding associated with diverticular disease comes from a perforated vasa recta located at the neck or apex of a diverticulum. The vasa recta penetrates the colonic wall from the serosa to the submucosa through obliquely oriented connective tissue septa. Protrusion of colonic mucosa through this connective tissue plane causes apposition of the diverticulum and the vasa recta (Fig. 65-4). Ulceration of the mucosa within the neck of the diverticulum and disruption of the arterial wall produces hemorrhage into the lumen of the bowel. Although diverticular disease is more prevalent in the left colon, right-sided lesions account for half or more episodes of bleeding.151,152 Risk factors for rebleeding are hypertension, NSAID use, coagulopathy, renal failure, ischemic heart disease, and cluster type diverticula.153,154

Figure 65-4. Colonoscopic view of a colonic diverticulum. A vasum rectum is seen entering the diverticulum and forming one of the walls.

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

May 5, 2017 | Posted by in GENERAL SURGERY | Comments Off on Acute Gastrointestinal Hemorrhage

Full access? Get Clinical Tree

Get Clinical Tree app for offline access